CN116110612A

CN116110612A - Intelligent consultation inquiry feedback processing method and system based on doctor-patient interaction

Info

Publication number: CN116110612A
Application number: CN202310122123.1A
Authority: CN
Inventors: 龙洋; 胡小玲
Original assignee: Chongzhou Traditional Chinese Medicine Hospital
Current assignee: Chongzhou Traditional Chinese Medicine Hospital
Priority date: 2023-02-15
Filing date: 2023-02-15
Publication date: 2023-05-12

Abstract

The embodiment of the application discloses an intelligent diagnosis-guiding query feedback processing method and system based on doctor-patient interaction, which are used for carrying out feature relevance mining between query interaction text and query disease knowledge point sets of an intelligent diagnosis-guiding query activity of a target doctor-patient interaction process, carrying out diagnosis-guiding content information feedback on a patient end corresponding to the target doctor-patient interaction process based on feature relevance mining results, acquiring diagnosis evaluation data of the patient end aiming at the fed back diagnosis-guiding content information, carrying out patient emotion analysis on the diagnosis evaluation data, and carrying out use feedback on a doctor-patient interaction platform operated by a doctor-patient service system according to patient emotion analysis results, so that the intelligent diagnosis-guiding feedback is carried out by combining the feature relevance between the query interaction text and the query disease knowledge point sets, the reliability of diagnosis-guiding feedback can be improved, and further carrying out use feedback after tracking the emotion of a subsequent patient, so that the reference information developed by the subsequent doctor-patient interaction platform can be conveniently provided.

Description

Intelligent consultation inquiry feedback processing method and system based on doctor-patient interaction

Technical Field

The application relates to the technical field of intelligent medical services, in particular to an intelligent consultation inquiry feedback processing method and system based on doctor-patient interaction.

Background

The rehabilitation therapy (such as needling therapy) is a therapy mode for promoting physical and mental dysfunction or disability caused by factors such as injury, diseases, development defects and the like to be normal or nearly normal, is an important component of rehabilitation medicine, can solve the problems of pain and the like caused by musculoskeletal system diseases, and has remarkable curative effects on consciousness disturbance, paresthesia, cognitive disturbance and the like caused by nervous system diseases.

As rehabilitation therapy becomes more widely used in the treatment of stroke patients and other severe patients, medical workers encounter new problems. In the diagnosis guiding stage before rehabilitation therapy, patients with current consciousness disturbance, cognitive disturbance, unsmooth speech communication, emotion control disturbance and paresthesia cannot be checked or prompted by themselves, and good communication is difficult to achieve, so that an intelligent diagnosis guiding query flow is required to be applied to conduct diagnosis guiding query on relevant guardians of the patients, and communication time before formal rehabilitation therapy is shortened. However, the guided diagnosis feedback in the related art is not reliable and lacks a follow-up patient use feedback tracking procedure.

Disclosure of Invention

The application provides an intelligent consultation inquiry feedback processing method and system based on doctor-patient interaction.

In a first aspect, an embodiment of the present application provides an intelligent diagnosis-guiding query feedback processing method based on doctor-patient interaction, which is applied to a doctor-patient service system, including:

feature relevance mining between query interaction text and query disease knowledge point sets is carried out on intelligent guide query activities of the target doctor-patient interaction flow, and guide diagnosis content information feedback is carried out on a patient end corresponding to the target doctor-patient interaction flow based on feature relevance mining results;

acquiring diagnosis evaluation data of the patient side aiming at the feedback diagnosis guiding content information;

and carrying out patient emotion analysis on the diagnosis evaluation data, and carrying out use feedback on a doctor-patient interaction platform operated by the doctor-patient service system according to the patient emotion analysis result.

In a possible implementation manner of the first aspect, the step of performing feature relevance mining between query interaction text and a query disease knowledge point set for the intelligent diagnosis-guiding query activity of the target doctor-patient interaction procedure, and performing diagnosis-guiding content information feedback on a patient end corresponding to the target doctor-patient interaction procedure based on a feature relevance mining result includes:

Acquiring a plurality of query interaction texts in an intelligent guide query activity of a target doctor-patient interaction flow, and respectively analyzing interaction keyword vectors of each query interaction text in the plurality of query interaction texts;

acquiring a query disease knowledge point set corresponding to the intelligent guided query activity, and respectively analyzing medical knowledge characteristics of each query disease knowledge point in the query disease knowledge point set;

splicing the interaction keyword vectors of the plurality of inquiry interaction texts and the corresponding medical knowledge features in the inquiry disease knowledge point set to generate a spliced feature vector sequence;

determining a first context characteristic of query interaction text corresponding to each interaction keyword vector in the intelligent guided query activity, and determining a second context characteristic of query disease knowledge points corresponding to each medical knowledge characteristic in the query disease knowledge point set;

combining the first context feature corresponding to each interaction keyword vector and the second context feature corresponding to each medical knowledge feature, performing feature node allocation on each spliced feature vector in the spliced feature vector sequence, and generating a target spliced feature vector sequence;

Combining the target spliced feature vector sequence to generate a diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity and at least one query feedback expansion result corresponding to the diagnosis guiding feedback result sequence;

carrying out diagnosis guiding content information feedback on a patient end corresponding to the target doctor-patient interaction flow by combining the diagnosis guiding feedback result and/or the query feedback expansion result;

combining the target spliced feature vector sequence to generate a diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding inquiry activity, wherein the diagnosis guiding feedback result sequence comprises the following steps:

loading the target spliced feature vector sequence to a guided diagnosis feedback prediction model, and outputting a first guided diagnosis feedback result of the guided diagnosis feedback result sequence;

generating medical knowledge features of a first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence;

combining the medical knowledge characteristics of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence and the carrying state of the first diagnosis guiding feedback result in the diagnosis guiding feedback result sequence, and updating the target spliced characteristic vector sequence based on characteristic fusion;

loading the updated target spliced feature vector sequence to a guided diagnosis feedback prediction model, and traversing the target spliced feature vector sequence based on the medical knowledge features of each acquired guided diagnosis feedback result until all guided diagnosis feedback sentences of the guided diagnosis feedback result sequence are acquired.

In a possible implementation manner of the first aspect, the guided diagnosis feedback prediction model includes a semantic editing unit and a feedback prediction unit, where loading the target stitching feature vector sequence into the guided diagnosis feedback prediction model, and outputting a first guided diagnosis feedback result of the guided diagnosis feedback result sequence includes:

loading the target spliced feature vector sequence to a semantic editing unit, and outputting a semantic editing vector of a first diagnosis guiding feedback result corresponding to the diagnosis guiding feedback result sequence;

analyzing a semantic editing vector of a first guided diagnosis feedback result corresponding to a guided diagnosis feedback result sequence through a feedback prediction unit to obtain first feedback prediction information, wherein the first feedback prediction information comprises the support degree of each guided diagnosis feedback result corresponding to the guided diagnosis feedback result sequence;

and combining the first feedback prediction information to determine a first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence.

In a possible implementation manner of the first aspect, in combination with the medical knowledge feature of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence and the carrying state of the medical knowledge feature in the diagnosis guiding feedback result sequence, updating the target spliced feature vector sequence based on feature fusion includes:

Splicing the interaction keyword vectors of the plurality of inquiry interaction texts, the medical knowledge features of the inquiry disease knowledge point set and the medical knowledge features of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence to update the spliced feature vector sequence;

and carrying out feature node allocation on each spliced feature vector in the updated spliced feature vector sequence by combining the first context feature corresponding to each interactive keyword vector, the second context feature corresponding to each medical knowledge feature of the query disease knowledge point set and the second context feature of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence so as to update the target spliced feature vector sequence.

In a possible implementation manner of the first aspect, the method further includes:

when all diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, determining a semantic editing vector from a plurality of semantic editing vectors generated by a semantic editing unit as an intelligent diagnosis guide inquiry editing vector and a semantic editing vector as inquiry disease editing vectors;

determining a correlation between the intelligent lead query edit vector and the query disease edit vector;

Determining a correlation result of the intelligent guided consultation inquiry activity and the inquiry disease knowledge point set by combining the correlation degree;

wherein one semantic editing vector as an intelligent guided query editing vector is a semantic editing vector corresponding to a first query interaction behavior feature distributed before an interaction keyword vector of the plurality of query interaction texts, and one semantic editing vector as a query disease editing vector is a semantic editing vector corresponding to a second query interaction behavior feature distributed between an interaction keyword vector of the plurality of query interaction texts and each medical knowledge feature of the query disease knowledge point set.

In a possible implementation manner of the first aspect, the determining, in combination with the first feedback prediction information, a first lead feedback result of the lead feedback result sequence includes:

in the first feedback prediction information, the support degrees are arranged in a descending order;

determining the first N supporters in a descending order queuing, taking the diagnosis guiding feedback results corresponding to the N supporters as reference diagnosis guiding feedback results of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence, loading the updated target spliced feature vector sequence to the semantic editing unit, traversing the target spliced feature vector sequence based on the acquired medical knowledge features of each diagnosis guiding feedback result until all diagnosis guiding feedback sentences of the diagnosis guiding feedback result sequence are acquired, and comprising the following steps:

Combining the reference diagnosis guiding feedback results of the first diagnosis guiding feedback result to generate the reference diagnosis guiding feedback results of the rest diagnosis guiding feedback results one by one;

combining the reference diagnosis guiding feedback results of all the diagnosis guiding feedback results in the diagnosis guiding feedback result sequences to determine N diagnosis guiding feedback result sequences;

the method further comprises the steps of:

for each of the N triage feedback result sequences, the following steps are performed respectively:

when all diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, determining a semantic editing vector from a plurality of semantic editing vectors generated by the semantic editing unit as an inquiry disease editing vector and a semantic editing vector as a diagnosis guide feedback editing vector;

and determining a correlation between the query disease edit vector and the lead feedback edit vector;

and when the maximum correlation degree is greater than the threshold correlation degree, determining the diagnosis guiding feedback result sequence corresponding to the correlation degree as the diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity, otherwise, determining that the diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity is not available.

when all diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, determining a semantic editing vector from a plurality of semantic editing vectors generated by the semantic editing unit as an intelligent diagnosis guide inquiry editing vector, a semantic editing vector as an inquiry disease editing vector and a semantic editing vector as a diagnosis guide feedback editing vector;

determining the correlation between the intelligent guided diagnosis inquiry edit vector and the inquiry disease edit vector, and determining the correlation between the intelligent guided diagnosis inquiry edit vector and the guided diagnosis feedback edit vector;

and when the correlation degree between the intelligent guided diagnosis inquiry edit vector and the inquiry disease edit vector is smaller than a first set correlation degree and the correlation degree between the intelligent guided diagnosis inquiry edit vector and the guided diagnosis feedback edit vector is larger than a second set correlation degree, determining the guided diagnosis feedback result sequence as a guided diagnosis feedback result sequence which is obtained only based on intelligent guided diagnosis inquiry activities.

In a possible implementation manner of the first aspect, the acquiring the interaction keyword vector and the medical knowledge feature and the feature node allocation are implemented based on a feature node allocation model, the method further comprises:

Training the feature node allocation model, the semantic editing unit and the feedback prediction unit in combination with a first lead feedback learning data sequence, wherein the first lead feedback learning data sequence comprises a plurality of first lead feedback learning data, each first lead feedback learning data comprises a first lead query activity to be learned, a first training query disease knowledge point set corresponding to the first lead query activity to be learned, and an a priori lead feedback result sequence corresponding to the first lead query activity and the first training query disease knowledge point set, wherein the feature node allocation model, the semantic editing unit and the feedback prediction unit are trained in combination with the first lead feedback learning data sequence, comprising:

in at least one first triage feedback learning data in the first triage feedback learning data sequence, for each first triage feedback learning data, performing the following steps:

acquiring a plurality of first training query interactive texts from a first to-be-studied consultation inquiring activity of the first consultation feedback learning data, and respectively analyzing interaction keyword vectors of each training query interactive text in the plurality of first training query interactive texts;

Acquiring a first training query disease knowledge point set corresponding to the first study guide query activity, and respectively analyzing medical knowledge features of each query disease knowledge point in the first training query disease knowledge point set, wherein each medical knowledge feature is consistent with the disease semantic direction of each interaction keyword vector;

converting at least one diagnosis guiding feedback result in a priori diagnosis guiding feedback result sequence into a diagnosis guiding feedback observation vector, generating a diagnosis guiding feedback observation vector set, and respectively analyzing the medical knowledge characteristics of each query disease knowledge point in the diagnosis guiding feedback observation vector set, wherein the medical knowledge characteristics of each query disease knowledge point in the diagnosis guiding feedback observation vector set are consistent with the disease semantic direction of each interaction keyword vector;

the interactive keyword vectors of the first training query interactive texts, the medical knowledge features of the first training query disease knowledge point set and the medical knowledge features of each query disease knowledge point in the guide feedback observation vector set are spliced to generate a first training spliced feature vector sequence;

determining first context characteristics of query interaction texts corresponding to interaction keyword vectors of each first training query interaction text in the first query activity to be studied, determining second context characteristics of query disease knowledge points corresponding to medical knowledge characteristics in the first training query disease knowledge point set, and determining second context characteristics of query disease knowledge points corresponding to medical knowledge characteristics in the guide feedback observation vector set;

Combining the first context features corresponding to the interaction keyword vectors and the second context features corresponding to the medical knowledge features, performing feature node allocation on each spliced feature vector in the first training spliced feature vector sequence, and generating a first training target spliced feature vector sequence;

combining the first training target spliced feature vector sequence to generate at least one diagnosis guiding feedback observation vector in the diagnosis guiding feedback observation vector set;

acquiring a learning cost value loss1 between the at least one diagnosis guiding feedback observation vector and an actual diagnosis guiding feedback result;

and training the characteristic node allocation model, the semantic editing unit and the feedback prediction unit at least by combining the learning cost value loss 1.

In a possible implementation manner of the first aspect, the training the feature node allocation model, the semantic editing unit and the feedback prediction unit in combination with the first guided feedback learning data sequence includes:

determining one semantic editing vector as an intelligent consultation inquiry editing vector, one semantic editing vector as an inquiry disease editing vector and one semantic editing vector as a consultation feedback editing vector from a plurality of semantic editing vectors generated by the semantic editing unit;

Combining the correlation degree between the intelligent guide inquiry edit vector and the corresponding inquiry disease edit vector and the correlation degree between the intelligent guide inquiry edit vector and the non-corresponding inquiry disease edit vector, and the correlation degree between the inquiry disease edit vector and the corresponding intelligent guide inquiry edit vector and the correlation degree between the inquiry disease edit vector and the non-corresponding intelligent guide inquiry edit vector, and obtaining a learning cost value loss2;

combining the correlation degree between the intelligent guided diagnosis inquiry edit vector and the corresponding guided diagnosis feedback edit vector and the correlation degree between the intelligent guided diagnosis inquiry edit vector and the non-corresponding guided diagnosis feedback edit vector and the correlation degree between the guided diagnosis feedback edit vector and the corresponding intelligent guided diagnosis inquiry edit vector and the correlation degree between the guided diagnosis feedback edit vector and the non-corresponding intelligent guided diagnosis inquiry edit vector to obtain a learning cost value loss3;

combining the correlation between the inquiry disease editing vector and the corresponding diagnosis guide feedback editing vector and the correlation between the inquiry disease editing vector and the non-corresponding diagnosis guide feedback editing vector, and the correlation between the diagnosis guide feedback editing vector and the corresponding inquiry disease editing vector and the correlation between the diagnosis guide feedback editing vector and the non-corresponding inquiry disease editing vector, and obtaining a learning cost value loss4;

And wherein training the feature node allocation model, the semantic editing unit, and the feedback prediction unit in combination with at least the learning cost value loss1, comprises:

and training the characteristic node allocation model, the semantic editing unit and the feedback prediction unit by combining the weighted fusion values of the learning cost value loss1, the learning cost value loss2, the learning cost value loss3 and the learning cost value loss 4.

In a second aspect, an embodiment of the present application provides a doctor-patient service system, including:

a processor;

and the memory is stored with a computer program which is executed to realize the intelligent consultation inquiry feedback processing method based on the doctor-patient interaction according to the first aspect.

Compared with the prior art, the intelligent diagnosis guiding and inquiring activity of the target doctor-patient interaction flow is subjected to feature relevance mining between an inquiring interaction text and an inquiring disease knowledge point set, diagnosis guiding content information feedback is carried out on a patient end corresponding to the target doctor-patient interaction flow based on feature relevance mining results, diagnosis guiding evaluation data of the patient end aiming at the fed-back diagnosis guiding content information are obtained, patient emotion analysis is carried out on the diagnosis guiding evaluation data, and use feedback is carried out on a doctor-patient interaction platform operated by the doctor-patient service system according to patient emotion analysis results, so that intelligent diagnosis guiding and inquiring feedback is carried out by combining feature relevance between the inquiring interaction text and the inquiring disease knowledge point set, diagnosis guiding feedback reliability can be improved, and further, use feedback is carried out after follow-up of emotion of a subsequent patient, so that reference information developed by the subsequent doctor-patient interaction platform can be conveniently provided.

Drawings

Fig. 1 is a schematic flow chart of steps of an intelligent consultation inquiry feedback processing method based on doctor-patient interaction according to an embodiment of the present application;

fig. 2 is a schematic block diagram of a doctor-patient service system for executing the intelligent consultation inquiry feedback processing method based on doctor-patient interaction in fig. 1 according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden in connection with the embodiments herein, are intended to be within the scope of the present application.

And 10, feature relevance mining between query interaction text and query disease knowledge point sets is carried out on intelligent diagnosis-guiding query activities of the target doctor-patient interaction flow, and diagnosis-guiding content information feedback is carried out on a patient end corresponding to the target doctor-patient interaction flow based on feature relevance mining results.

And step 20, acquiring the diagnosis evaluation data of the patient side aiming at the feedback diagnosis guiding content information.

For example, the diagnosis evaluation data may be voice evaluation data, text evaluation data, or the like of the patient-side diagnosis guiding content information.

And step 30, carrying out patient emotion analysis on the diagnosis evaluation data, and carrying out use feedback on a doctor-patient interaction platform operated by the doctor-patient service system according to the patient emotion analysis result.

In this embodiment, the emotion of the diagnosis evaluation data may be classified by using a relevant technology of machine learning based on an analysis method of a corpus, so as to generate an emotion analysis result of the patient, such as a feeling of Qi, a horror, a fear, an interest, a happiness, a sadness, and the like. The machine learning method generally needs to make the classification model learn the rule in the training data, and then predict the test data by using the trained model, thereby further.

Based on the above steps, in this embodiment, feature relevance mining is performed between an intelligent diagnosis-guiding query activity of a target doctor-patient interaction process and a query disease knowledge point set, diagnosis-guiding content information feedback is performed on a patient end corresponding to the target doctor-patient interaction process based on feature relevance mining results, diagnosis-guiding evaluation data of the patient end for the fed-back diagnosis-guiding content information is obtained, patient emotion analysis is performed on the diagnosis-guiding evaluation data, and feedback is performed on a doctor-patient interaction platform operated by the doctor-patient service system according to patient emotion analysis results, so that intelligent diagnosis-guiding feedback is performed by combining feature relevance between the query interaction text and the query disease knowledge point set, diagnosis-guiding feedback reliability can be improved, and further, follow-up patient emotion is tracked and then used to provide reference information developed by a follow-up doctor-patient interaction platform.

Step S100: and acquiring a plurality of query interaction texts in the intelligent guided query activity, and respectively analyzing interaction keyword vectors of each query interaction text in the plurality of query interaction texts.

In this embodiment, a smart lead query activity may be composed of a plurality of query interactive texts, where the query interactive texts are texts of lead queries (e.g., lead queries for rehabilitation therapy, etc.) initiated on the healthcare platform for the patient. Each query interaction text corresponds to a corresponding set of query disease knowledge points, and the sequence of query disease knowledge points comprises at least one query disease knowledge point used for marking the knowledge points of the text diseases of the query interaction text.

For a plurality of query interaction texts acquired in the intelligent guided query activity, a word vector encoding network can be called to extract an interaction keyword vector of each query interaction text, wherein the interaction keyword vector is used for representing a feature vector of the corresponding query interaction text. Each query interaction text corresponds to an interaction keyword vector.

For some possible embodiments, after the interaction keyword vector of each query interaction text is extracted through the word vector encoding network, based on continuity of each query text in the time sequence dimension in the intelligent query activity, in order to reduce the processing capacity, the extracted interaction keyword vector of each query interaction text may be integrated, for example, the relevance of the interaction keyword vectors of two query interaction texts continuous in the time sequence dimension may be obtained. And integrating the interaction keyword vectors with high correlation.

Step S200: acquiring a query disease knowledge point set corresponding to the intelligent guide query activity, and respectively analyzing the medical knowledge characteristics of each query disease knowledge point in the query disease knowledge point set.

As can be seen from the above step S100, each intelligent guided diagnosis query activity includes a corresponding query disease knowledge point set, and each guided diagnosis feedback result in the query disease knowledge point set is mapped to a medical knowledge feature in another value range, where each medical knowledge feature and each interaction keyword vector can be consistent with the disease semantic direction.

Step S300: and splicing the interaction keyword vectors of the query interaction texts and the corresponding medical knowledge features in the query disease knowledge point set to generate a spliced feature vector sequence.

The method provided by the embodiment of the application further comprises the process of obtaining the first query interaction behavior characteristics and the first query interaction behavior distribution corresponding to the first query interaction behavior, and obtaining the second query interaction behavior characteristics and the second query interaction behavior distribution corresponding to the second query interaction behavior.

In step S300, the first query interaction behavior feature, the interaction keyword vectors of the multiple query interaction texts, the second query interaction behavior feature and the medical knowledge feature of the query disease knowledge point set are spliced one by one to generate a spliced feature vector sequence.

Step S400: determining a first context characteristic of query interaction text corresponding to each interaction keyword vector in the intelligent guided query activity, and determining a second context characteristic of query disease knowledge points corresponding to each medical knowledge characteristic in the query disease knowledge point set.

Step S500: and carrying out feature node allocation on each spliced feature vector in the spliced feature vector sequence by combining the first context feature corresponding to each interactive keyword vector and the second context feature corresponding to each medical knowledge feature to generate a target spliced feature vector sequence.

In step S500, feature node allocation is performed on each spliced feature vector in the spliced feature vector sequence by combining the first query interaction behavior distribution, the first context feature corresponding to each determined interaction keyword vector, the second query interaction behavior distribution, and the second context feature corresponding to each medical knowledge feature, so that the position embedding process of the spliced feature vector is completed, and a target spliced feature vector sequence is generated.

Step S600: and combining the target spliced feature vector sequence to generate a diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity and at least one query feedback expansion result corresponding to the diagnosis guiding feedback result sequence.

For some possible embodiments, generating a diagnosis-guiding feedback result sequence corresponding to the intelligent diagnosis-guiding query activity in combination with the target spliced feature vector sequence may specifically include:

step S601: and loading the target spliced feature vector sequence to a guided diagnosis feedback prediction model, and outputting the current guided diagnosis feedback result of the guided diagnosis feedback result sequence. The current diagnosis guiding feedback result of the obtained diagnosis guiding feedback result sequence is the first diagnosis guiding feedback result.

For some possible embodiments, the current lead feedback result of the lead feedback result sequence may be obtained based on the following manner. The guided diagnosis feedback prediction model comprises a semantic editing unit and a feedback prediction unit, wherein a target spliced feature vector sequence is loaded to the semantic editing unit to obtain a semantic editing vector of a current guided diagnosis feedback result corresponding to the guided diagnosis feedback result sequence, and encoding of the current guided diagnosis feedback result is completed. And then, combining a feedback prediction unit to analyze the semantic editing vector of the current guided diagnosis feedback result corresponding to the guided diagnosis feedback result sequence to obtain first feedback prediction information, and finishing decoding, wherein the support degree of each guided diagnosis feedback result corresponding to the guided diagnosis feedback result sequence is covered in the first feedback prediction information. And then determining the current lead diagnosis feedback result of the lead diagnosis feedback result sequence by combining the first feedback prediction information. For example, determining a feedback statement with the largest support degree in the first feedback prediction information, and taking the leading feedback result corresponding to the feedback statement as the current leading feedback result of the leading feedback result sequence. At this time, a diagnosis-guiding feedback result sequence corresponding to the intelligent diagnosis-guiding inquiry activity is finally generated. In addition, a plurality of feedback sentences with larger support degree can be determined in the first feedback prediction information, and a plurality of diagnosis guiding feedback results corresponding to the feedback sentences are all determined to be the alternative diagnosis guiding feedback results of the diagnosis guiding feedback result sequence. At this time, a plurality of diagnosis-guiding feedback result sequences corresponding to the intelligent diagnosis-guiding inquiry activities are finally generated.

For example, in combination with the first feedback prediction information, determining the first lead feedback result of the lead feedback result sequence includes: in the first feedback prediction information, the support degrees are arranged in a descending order; determining the first N supporters in the descending order queue, and taking the diagnosis guiding feedback results corresponding to the N supporters as the reference diagnosis guiding feedback result of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence, wherein N is more than or equal to 1. Loading the updated target spliced feature vector sequence to a semantic editing unit, and iterating until all diagnosis guiding feedback sentences of a diagnosis guiding feedback result sequence are obtained, wherein the method comprises the following steps: combining the reference diagnosis guiding feedback results of the first diagnosis guiding feedback result to generate the reference diagnosis guiding feedback results of the rest diagnosis guiding feedback results one by one; and determining N guide diagnosis feedback result sequences by combining the reference guide diagnosis feedback results of the guide diagnosis feedback results in the guide diagnosis feedback result sequences. In other words, each reference diagnosis guiding feedback result of the first diagnosis guiding feedback result is respectively loaded to the diagnosis guiding feedback prediction model, and N reference diagnosis guiding feedback results of the second diagnosis guiding feedback result are obtained. Thus, N2 first diagnosis guiding feedback results and a reference diagnosis guiding feedback result of the second diagnosis guiding feedback result are obtained.

Step S602: judging whether the current lead diagnosis feedback result of the lead diagnosis feedback result sequence obtained in the step S601 is all lead diagnosis feedback sentences. If not, step S603 is performed; if so, ending.

Step S603: and generating medical knowledge characteristics of the current diagnosis guiding feedback result of the diagnosis guiding feedback result sequence.

Step S604: and updating the target spliced feature vector sequence based on feature fusion by combining the medical knowledge features of the current guided diagnosis feedback result of the guided diagnosis feedback result sequence and the carrying state of the current guided diagnosis feedback result in the guided diagnosis feedback result sequence.

For example, the current lead diagnosis feedback result of the lead diagnosis feedback result sequence is the first lead diagnosis feedback result (the processing process of the lead diagnosis feedback results of other positions is the same), and the target spliced feature vector sequence is updated based on feature fusion by combining the medical knowledge features of the first lead diagnosis feedback result of the lead diagnosis feedback result sequence and the carrying state thereof in the lead diagnosis feedback result sequence, including: and splicing the interactive keyword vectors of the query interactive texts, the medical knowledge features of the query disease knowledge point set and the medical knowledge features of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence to update the spliced feature vector sequence. And then, combining the first context feature corresponding to each interaction keyword vector, the second context feature corresponding to each medical knowledge feature of the query disease knowledge point set and the second context feature of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence, and performing feature node allocation on each spliced feature vector in the updated spliced feature vector sequence to update the target spliced feature vector sequence so as to finish updating.

For other possible embodiments, combining the medical knowledge features of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence and the carrying state thereof in the diagnosis guiding feedback result sequence, updating the target spliced feature vector sequence based on feature fusion, including: and splicing the interactive keyword vectors of the query interactive texts, the medical knowledge features of the query disease knowledge point set and the medical knowledge features of the first diagnosis guiding feedback result of the diagnosis guiding feedback result sequence to update the spliced feature vector sequence. And then, combining the individual corresponding to each interactive keyword vector, the individual corresponding to each medical knowledge feature of the query disease knowledge point set and the individual corresponding to the query disease knowledge point of the first guide feedback result of the guide feedback result sequence, and performing individual embedded coding on each spliced feature vector in the spliced feature vector sequence to update the spliced feature vector sequence to finish updating. Or combining the individual corresponding to each interactive keyword vector, the individual corresponding to each medical knowledge feature of the query disease knowledge point set and the individual corresponding to the query disease knowledge point of the first guide feedback result of the guide feedback result sequence, and performing individual embedded coding on each spliced feature vector in the target spliced feature vector sequence to update the target spliced feature vector sequence.

And returning to the step S601, loading the updated target spliced feature vector sequence to a guided diagnosis feedback prediction model, and traversing the target spliced feature vector sequence based on the acquired medical knowledge features of each guided diagnosis feedback result until all guided diagnosis feedback sentences of the guided diagnosis feedback result sequence are acquired.

When all diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, a semantic editing vector is determined from a plurality of semantic editing vectors generated by the semantic editing unit to serve as an intelligent diagnosis guide inquiry editing vector and a semantic editing vector is determined to serve as an inquiry disease editing vector. For example, a semantic editing vector as an intelligent guided query editing vector may be a semantic editing vector corresponding to a first query interactive behavior feature distributed before an interactive keyword vector of a plurality of query interactive texts, and a semantic editing vector as a query disease editing vector is a semantic editing vector corresponding to a second query interactive behavior feature distributed between an interactive keyword vector of a plurality of query interactive texts and each medical knowledge feature of a set of query disease knowledge points. Then, a correlation between the intelligent lead query edit vector and the query disease edit vector is determined. In some embodiments, a cosine distance may be used to obtain a correlation between the intelligent lead query edit vector and the query disease edit vector. At this time, the correlation is located at [ -1,1], and the closer the correlation is to 1, the higher the correlation is. And then determining the correlation result of the intelligent guided consultation inquiry activity and the inquiry disease knowledge point set by combining the correlation degree. Furthermore, the trend correlation degree of the acquired diagnosis guiding feedback result sequence and the query disease knowledge point set can be further evaluated. When a plurality of diagnosis guiding feedback result sequences corresponding to intelligent diagnosis guiding inquiry activities are acquired, the acquired plurality of diagnosis guiding feedback result sequences are screened based on the correlation degree of the generated diagnosis guiding feedback result sequences and the inquiry disease knowledge point set, and the diagnosis guiding feedback result sequences with low correlation degree are cleaned.

For example, for each of the N triage feedback result sequences, the following steps are performed: when all the diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, a semantic editing vector is determined from a plurality of semantic editing vectors generated by the semantic editing unit to serve as an inquiry disease editing vector and a semantic editing vector is determined to serve as a diagnosis guide feedback editing vector. For example, one semantic editing vector as the query disease editing vector may be a semantic editing vector corresponding to a second query interaction behavior feature distributed between the interaction keyword vector of the plurality of query interaction texts and each of the medical knowledge features of the query disease knowledge point set, and one semantic editing vector as the lead feedback editing vector is a semantic editing vector corresponding to a second query interaction behavior feature distributed between each of the medical knowledge features of the query disease knowledge point set and each of the medical knowledge features of the lead feedback result sequence. Then, a correlation between the query disease edit vector and the lead feedback edit vector is determined. In some embodiments, a vector cosine distance is used to obtain a correlation between the query disease edit vector and the lead feedback edit vector. And when the maximum correlation degree is greater than the threshold correlation degree, determining that the diagnosis guiding feedback result sequence corresponding to the correlation degree is the diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity, otherwise, determining that the diagnosis guiding feedback result sequence corresponding to the intelligent diagnosis guiding query activity does not exist.

The higher the correlation degree is, the higher the accuracy is, the correlation degree between the query disease knowledge point set and the diagnosis-guiding feedback result sequence is represented, and the prior diagnosis-guiding feedback result sequence and the false diagnosis-guiding feedback result sequence are stimulated. And cleaning out the diagnosis guide feedback result sequence with low correlation degree based on the diagnosis guide feedback result sequence generated by the correlation degree screening of the query disease knowledge point set and the diagnosis guide feedback result sequence, so that the accuracy of the diagnosis guide feedback result sequence can be increased.

In addition, the diagnosis guiding feedback result sequence obtained only through the intelligent diagnosis guiding inquiry activity can be determined based on the correlation degree of the intelligent diagnosis guiding inquiry activity and the diagnosis guiding feedback result sequence and the correlation degree of the inquiry disease knowledge point set and the diagnosis guiding feedback result sequence. Specifically, for each of the N triage feedback result sequences, the following steps are performed: when all the diagnosis guide feedback sentences of the diagnosis guide feedback result sequence are obtained, a semantic editing vector is determined from a plurality of semantic editing vectors generated by the semantic editing unit to serve as an intelligent diagnosis guide inquiry editing vector, one semantic editing vector serves as an inquiry disease editing vector and one semantic editing vector serves as a diagnosis guide feedback editing vector. For example, one semantic editing vector as an intelligent lead query edit vector may be a semantic editing vector corresponding to a first query interaction behavior feature distributed before an interaction keyword vector of a plurality of query interaction texts, one semantic editing vector as a query disease edit vector may be a semantic editing vector corresponding to a second query interaction behavior feature distributed between an interaction keyword vector of a plurality of query interaction texts and each medical knowledge feature of a query disease knowledge point set, and one semantic editing vector as a lead feedback edit vector may be a semantic editing vector corresponding to a second query interaction behavior feature distributed between each medical knowledge feature of a query disease knowledge point set and each medical knowledge feature of a lead feedback result sequence. And then determining the correlation degree between the intelligent guided diagnosis inquiry edit vector and the inquiry disease edit vector, and determining the correlation degree between the intelligent guided diagnosis inquiry edit vector and the guided diagnosis feedback edit vector. For example, the correlation degree between the intelligent guided diagnosis inquiry activity and the guided diagnosis feedback result sequence is obtained based on the cosine distance, and the correlation degree between the inquiry disease editing vector and the guided diagnosis feedback editing vector is obtained. And when the correlation degree between the intelligent guided diagnosis inquiry edit vector and the inquiry disease edit vector is smaller than the first set correlation degree and the intelligent guided diagnosis inquiry edit vector and the guided diagnosis feedback edit vector are larger than the second set correlation degree, determining the guided diagnosis feedback result sequence as a guided diagnosis feedback result sequence which is only obtained based on the intelligent guided diagnosis inquiry activity.

Step S700: and carrying out diagnosis guiding content information feedback on the patient end corresponding to the target doctor-patient interaction flow by combining the diagnosis guiding feedback result and/or the query feedback expansion result.

The training method of each module used in the generation process of the diagnosis guiding feedback result sequence in the embodiment of the application is described below. The feature node distribution model, the semantic editing unit and the feedback prediction unit are trained by combining a first guided diagnosis feedback learning data sequence, wherein the first guided diagnosis feedback learning data sequence comprises a plurality of first guided diagnosis feedback learning data, and each first guided diagnosis feedback learning data comprises a first guided diagnosis query activity, a first training query disease knowledge point set corresponding to the first guided diagnosis query activity and a priori guided diagnosis feedback result sequence corresponding to the first intelligent guided diagnosis query activity and the first query disease knowledge point set.

(1) And acquiring a plurality of first training query interactive texts in a first to-be-learned guide query activity of the first guide query feedback learning data, and respectively analyzing the interaction keyword vectors of each training query interactive text in the plurality of first training query interactive texts.

(2) And acquiring a first training disease knowledge point set corresponding to the first study consultation-guiding query activity, and respectively analyzing the medical knowledge characteristics of each disease knowledge point in the first training disease knowledge point set, wherein each medical knowledge characteristic is consistent with the disease semantic direction which can be achieved by each interaction keyword vector.

(3) At least one diagnosis guiding feedback result in the prior diagnosis guiding feedback result sequence is converted into a diagnosis guiding feedback observation vector, a diagnosis guiding feedback observation vector set is generated, and medical knowledge features of each query disease knowledge point in the diagnosis guiding feedback observation vector set are respectively analyzed, wherein the medical knowledge features of each query disease knowledge point in the diagnosis guiding feedback observation vector set are consistent with the disease semantic direction which can be achieved by each interaction keyword vector.

(4) And splicing the interaction keyword vectors of the plurality of first training query interaction texts, the medical knowledge features of the first training query disease knowledge point set and the medical knowledge features of each query disease knowledge point in the guide feedback observation vector set to generate a first training spliced feature vector sequence.

(5) Determining first context characteristics of query interaction texts corresponding to interaction keyword vectors of each first training query interaction text in a first query activity to be studied, determining second context characteristics of query disease knowledge points corresponding to medical knowledge characteristics in a first training query disease knowledge point set, and determining second context characteristics of query disease knowledge points corresponding to medical knowledge characteristics in a guide feedback observation vector set.

(6) And carrying out feature node allocation on each spliced feature vector in the first training spliced feature vector sequence by combining the first context feature corresponding to each interaction keyword vector and the second context feature corresponding to each medical knowledge feature to generate a first training target spliced feature vector sequence.

(7) And combining the first training target spliced feature vector sequence to generate at least one diagnosis guiding feedback observation vector in the diagnosis guiding feedback observation vector set.

(8) And acquiring a learning cost value loss1 between at least one diagnosis guiding feedback observation vector and an actual diagnosis guiding feedback result.

(9) And training a feature node distribution model, a semantic editing unit and a feedback prediction unit at least by combining the learning cost value loss1.

In the embodiment of the application, when the learning cost value is constructed, in addition to the learning cost value loss1 between the guided diagnosis feedback observation vector and the actual guided diagnosis feedback result, the multi-element (triplet) learning cost value between the intelligent guided diagnosis inquiry activity and the inquiry disease knowledge point set, the multi-element learning cost value between the intelligent guided diagnosis inquiry activity and the generated guided diagnosis feedback result sequence and the multi-element learning cost value between the inquiry disease knowledge point set and the generated guided diagnosis feedback result sequence are also introduced. The intelligent diagnosis guiding and inquiring activity, the disease inquiring knowledge point set and the diagnosis guiding and inquiring feedback result sequence are spliced feature vectors obtained based on the same diagnosis guiding and feedback prediction model, so that the multi-element learning cost value is introduced, and the purpose of the intelligent diagnosis guiding and inquiring activity or disease inquiring knowledge point set is to the greatest extent close to the diagnosis guiding and inquiring feedback result sequence generated by the diagnosis guiding and feedback prediction model in the dimension of the disease.

For example, the process of obtaining the multiple learning cost value may include: a semantic editing vector as an intelligent consultation inquiry editing vector, a semantic editing vector as an inquiry disease editing vector and a semantic editing vector as a consultation feedback editing vector are determined from a plurality of semantic editing vectors generated by a semantic editing unit. And then, combining the correlation degree between the intelligent guide inquiry edit vector and the corresponding inquiry disease edit vector and the correlation degree between the intelligent guide inquiry edit vector and the non-corresponding inquiry disease edit vector, and the correlation degree between the inquiry disease edit vector and the corresponding intelligent guide inquiry edit vector and the correlation degree between the inquiry disease edit vector and the non-corresponding intelligent guide inquiry edit vector to obtain a learning cost value loss2, and determining the learning cost value loss as a multi-element learning cost value between the intelligent guide inquiry activity and the inquiry disease knowledge point set. The manner of obtaining the learning cost value loss2 can refer to the following formula:

lossoHH2＝G1+G2

G1＝Max(H(V，K')+u-H(V，K),0)

G2＝Max(H(V'，K)+u-H(V，K),0)

where lossoHH2 is the learning cost value lossohhhh 2, V, K and K' are each the intelligent lead query activity, the a priori query disease knowledge point set (the query disease knowledge point set corresponding to the intelligent lead query activity), the fake query disease knowledge point set (the query disease knowledge point set not corresponding to the intelligent lead query activity, i.e., the query disease knowledge point set corresponding to the other intelligent lead query activities). K. V and V' are each a set of query disease knowledge points, a priori intelligent guided query activity (intelligent guided query activity corresponding to a set of query disease knowledge points), a false intelligent guided query activity (intelligent guided query activity not corresponding to a set of query disease knowledge points, i.e., intelligent guided query activity corresponding to other sets of query disease knowledge points). u is an adjustment parameter, which can be set according to practice, such as 0.3. When training, a batch of guide feedback learning data is generally tested, and multiple groups of intelligent guide inquiry activities, inquiry disease knowledge point sets and semantic editing vectors of guide feedback result sequences are obtained for the batch of guide feedback learning data, wherein the same groups of intelligent guide inquiry activities, inquiry disease knowledge point sets and semantic editing vectors of guide feedback result sequences correspond to each other, and the intelligent guide inquiry activities, inquiry disease knowledge point sets and semantic editing vectors of guide feedback result sequences of different groups do not correspond to each other. In a batch of diagnosis-guiding feedback learning data, if an intelligent diagnosis-guiding inquiry edit vector has a plurality of non-corresponding inquiry disease edit vectors, a mean value of the correlation degree between the intelligent diagnosis-guiding inquiry edit vector and each non-corresponding inquiry disease edit vector is obtained, and is determined as K (V, K'). If there are multiple non-corresponding intelligent lead inquiry edit vectors for an inquiry disease edit vector, the average of the correlation between the inquiry disease edit vector and each non-corresponding intelligent lead inquiry edit vector is obtained and determined as K (V', K).

And combining the correlation degree between the intelligent guided diagnosis inquiry edit vector and the corresponding guided diagnosis feedback edit vector, the correlation degree between the intelligent guided diagnosis inquiry edit vector and the non-corresponding guided diagnosis feedback edit vector, and the correlation degree between the guided diagnosis feedback edit vector and the corresponding intelligent guided diagnosis inquiry edit vector, the guided diagnosis feedback edit vector and the non-corresponding intelligent guided diagnosis inquiry edit vector to obtain a learning cost value lossoHH3, and taking the learning cost value as a multi-element learning cost value between the intelligent guided diagnosis inquiry activity and the guided diagnosis feedback result sequence.

loss3＝G3+G4

G3＝Max(H(V，M')+u-H(V，M)，0)

G4＝Max(H(V'，M)+u-H(V，M)，0)

Where loss3 is a learning cost value lossoHH3, V, M, and M' each represent an intelligent guided query activity, a priori guided feedback result sequence (a guided feedback result sequence corresponding to the intelligent guided query activity), and a false guided feedback result sequence (a guided feedback result sequence not corresponding to the intelligent guided query activity, i.e., a guided feedback result sequence corresponding to other intelligent guided query activities). M, V and V' each represent a guided diagnosis feedback result sequence, an a priori intelligent guided diagnosis query activity (an intelligent guided diagnosis query activity corresponding to the guided diagnosis feedback result sequence), a false intelligent guided diagnosis query activity (an intelligent guided diagnosis query activity not corresponding to the guided diagnosis feedback result sequence, i.e. an intelligent guided diagnosis query activity corresponding to other guided diagnosis feedback result sequences), and u is an adjustment coefficient.

And combining the correlation degree between the inquiry disease editing vector and the corresponding consultation feedback editing vector and the correlation degree between the inquiry disease editing vector and the non-corresponding consultation feedback editing vector, and the correlation degree between the consultation feedback editing vector and the corresponding consultation disease editing vector and the correlation degree between the consultation feedback editing vector and the non-corresponding consultation disease editing vector to obtain a learning cost value lossoHH4 as a multi-element learning cost value between the consultation disease knowledge point set and the consultation feedback result sequence.

loss4＝G5+G6

G5＝Max(H(K，M')+u-H(K，M)，0)

G6＝Max(H(K'，M)+u-H(K，M)，0)

The loss4 is a learning cost value lossoHH4, K, M and M' respectively represent a query disease knowledge point set, a priori guided diagnosis feedback result sequence (a guided diagnosis feedback result sequence corresponding to the query disease knowledge point set), and a false guided diagnosis feedback result sequence (a guided diagnosis feedback result sequence not corresponding to the query disease knowledge point set, namely, a guided diagnosis feedback result sequence corresponding to other query disease knowledge point sets). M, K and K' each represent a lead feedback result sequence, a priori set of query disease knowledge points (set of query disease knowledge points corresponding to lead feedback result sequences), a sham set of query disease knowledge points (set of query disease knowledge points not corresponding to lead feedback result sequences, i.e., set of query disease knowledge points corresponding to other lead feedback result sequences). u is the adjustment coefficient.

Training a feature node allocation model, a semantic editing unit and a feedback prediction unit at least in combination with a learning cost value lossoHH1, comprising: and training a feature node distribution model, a semantic editing unit and a feedback prediction unit by combining the weighted fusion values of the learning cost value lossoHH1, the learning cost value lossoHH2, the learning cost value lossoHH3 and the learning cost value lossoHH 4.

For some possible embodiments, the method further comprises a pre-training process before training the feature node assignment model, the semantic editing unit, and the feedback prediction unit in conjunction with the first lead feedback learning data sequence. The training process further includes: and training a feature node distribution model and a semantic editing unit based on a second guided diagnosis feedback learning data sequence, wherein the second guided diagnosis feedback learning data sequence comprises a plurality of second guided diagnosis feedback learning data, and each second guided diagnosis feedback learning data comprises a second guided diagnosis query activity to be learned and a second training query disease knowledge point set corresponding to the second guided diagnosis query activity to be learned. The second diagnosis guiding feedback learning data sequence is different from the first diagnosis guiding feedback learning data sequence in that: the second guided diagnosis feedback learning data sequence is a guided diagnosis feedback learning data sequence of calibration learning basis data, namely, corresponding labels are not matched so as to annotate prior information.

When the feature node distribution model and the semantic editing unit are trained based on the second diagnosis-guiding feedback learning data sequence, in at least one second diagnosis-guiding feedback learning data in the second diagnosis-guiding feedback learning data sequence, for each second diagnosis-guiding feedback learning data, performing the following steps:

(1) And acquiring a plurality of second training query interactive texts in the to-be-studied consultation inquiring activity of the second consultation feedback learning data, and respectively analyzing the interaction keyword vectors of each training query interactive text in the plurality of second training query interactive texts.

(2) And acquiring a second training disease knowledge point set corresponding to the second study consultation guiding and inquiring activity, and respectively analyzing the medical knowledge characteristics of each disease knowledge point in the second training disease knowledge point set, wherein each medical knowledge characteristic is consistent with the disease semantic direction which can be achieved by each interaction keyword vector.

(3) And splicing the interaction keyword vectors of the plurality of second training query interaction texts and the medical knowledge features of the second training query disease knowledge point set to generate a second training spliced feature vector sequence.

(4) And determining the first context characteristics of the query interactive text corresponding to the interactive keyword vector of each second training query interactive text in the second query activity to be studied, and determining the second context characteristics of the query disease knowledge points corresponding to each medical knowledge characteristic in the second training query disease knowledge point set.

(5) And carrying out feature node allocation on each spliced feature vector in a second training spliced feature vector sequence by combining the first context feature corresponding to each interaction keyword vector and the second context feature corresponding to each medical knowledge feature to generate a second training target spliced feature vector sequence.

And carrying out prepositive training on a second guided diagnosis feedback learning data sequence of a large number of guided diagnosis feedback result sequences without annotation information in advance, and learning the carrying state of the spliced feature vector of the intelligent guided diagnosis query activity data knowledge and the query disease knowledge point set, wherein the spliced feature vector of the intelligent guided diagnosis query activity is closer to the spliced feature vector of the query disease knowledge point set in tendency based on the carrying state. And then loading the second training target spliced feature vector sequence to a semantic editing unit, and determining one semantic editing vector serving as an intelligent consultation inquiry editing vector and one semantic editing vector serving as an inquiry disease editing vector from a plurality of semantic editing vectors generated by the semantic editing unit. And then, combining the correlation degree between the intelligent guide inquiry edit vector and the corresponding inquiry disease edit vector and the correlation degree between the intelligent guide inquiry edit vector and the non-corresponding inquiry disease edit vector, and the correlation degree between the inquiry disease edit vector and the corresponding intelligent guide inquiry edit vector and the correlation degree between the inquiry disease edit vector and the non-corresponding intelligent guide inquiry edit vector, and obtaining a learning cost value loss5. Next, the feature node allocation model and the semantic editing unit are trained in conjunction with the learning cost value loss5.

In connection with the same inventive concept, and as shown in fig. 2, the embodiment of the present application further provides a doctor-patient service system, where the doctor-patient service system 100 may have a relatively large difference due to different configurations or performances, and may include one or more central processing units (Central Processing Units, CPU) 112 (e.g., one or more processors) and a memory 111. Wherein the memory 111 may be a transient storage or a persistent storage. The program stored in the memory 111 may include one or more modules, each of which may include a series of instruction operations in the doctor-patient service system 100. Still further, the central processor 112 may be configured to communicate with the memory 111 to execute a series of instruction operations in the memory 111 on the doctor-patient service system 100.

The doctor-patient service system 100 may also include one or more power supplies, one or more communication units 113, one or more delivery to output interfaces, and/or one or more operating systems, such as Windows ServerTM, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.

In addition, the embodiment of the application also provides a storage medium for storing a computer program for executing the method provided by the embodiment.

The present embodiments also provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform the method provided by the above embodiments.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, where the above program may be stored in a computer readable storage medium, and when the program is executed, the program performs steps including the above method embodiments; and the aforementioned storage medium may be at least one of the following media: read-only Memory (ROM), RAM, magnetic disk or optical disk, etc.

It should be noted that, in the present specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are referred to each other, and each embodiment is intended to describe a difference from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, with reference to the description of the method embodiments in part. The apparatus and system embodiments described above are merely illustrative, in which elements illustrated as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

The foregoing is merely one specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An intelligent consultation inquiry feedback processing method based on doctor-patient interaction is applied to the doctor-patient service system, and is characterized by comprising the following steps:

2. The method for intelligent diagnosis-guiding query feedback processing based on doctor-patient interaction according to claim 1, wherein the step of performing feature relevance mining between query interaction text and query disease knowledge point set for the intelligent diagnosis-guiding query activity of the target doctor-patient interaction flow and performing diagnosis-guiding content information feedback on the patient end corresponding to the target doctor-patient interaction flow based on feature relevance mining results comprises the steps of:

3. The doctor-patient interaction-based intelligent diagnosis inquiry feedback processing method according to claim 2, wherein the diagnosis-guiding feedback prediction model includes a semantic editing unit and a feedback prediction unit, wherein loading the target spliced feature vector sequence into the diagnosis-guiding feedback prediction model, and outputting a first diagnosis-guiding feedback result of the diagnosis-guiding feedback result sequence includes:

4. The doctor-patient interaction-based intelligent consultation inquiry feedback processing method according to claim 2, wherein the updating the target spliced feature vector sequence based on feature fusion by combining the medical knowledge feature of the first consultation feedback result of the consultation feedback result sequence and the carrying state of the medical knowledge feature in the consultation feedback result sequence comprises:

5. The doctor-patient interaction-based intelligent consultation inquiry feedback processing method according to claim 2, further comprising:

6. The doctor-patient interaction-based intelligent consultation inquiry feedback processing method according to claim 3, wherein the determining the first consultation feedback result of the consultation feedback result sequence by combining the first feedback prediction information includes:

the method further comprises the steps of:

7. The doctor-patient interaction-based intelligent consultation inquiry feedback processing method of claim 6, further comprising:

8. The doctor-patient interaction-based intelligent consultation inquiry feedback processing method of claim 3, wherein obtaining the interaction keyword vector and the medical knowledge feature and the feature node allocation are implemented based on a feature node allocation model, the method further comprising:

9. The doctor-patient interaction-based intelligent lead query feedback processing method as claimed in claim 8, wherein the training the feature node allocation model, the semantic editing unit, and the feedback prediction unit in combination with the first lead feedback learning data sequence includes:

10. A doctor-patient service system, comprising:

a processor;

a memory, wherein the memory stores a computer program, and the computer program is executed to implement the intelligent diagnosis-guiding query feedback processing method based on doctor-patient interaction as set forth in any one of claims 1 to 9.